topic Regression Technique in Statistical Procedures

Regression Technique

buder — Wed, 08 Mar 2017 21:06:21 GMT

If your dependent variable is a continuous variable but is bounded between 0 and 100, what type of regression analysis would work best?

Re: Regression Technique

PGStats — Wed, 08 Mar 2017 21:45:02 GMT

Assuming that your dependent variable is a percentage, you should consider transforming it for doing regression. Good documentation on this topic is given in SAS/IML doc :

http://support.sas.com/documentation/cdl/en/imlsug/66113/HTML/default/viewer.htm#imlsug_ugvartransform_sect007.htm

arcsin(sqrt(y)) (where y is between 0 and 1) is the preferred transformation when the denominators of the proportions are within the same order of magnitude.

Re: Regression Technique

buder — Wed, 08 Mar 2017 21:47:21 GMT

Thanks! The depdenent variable is not a percentage, it's a ranking (0 is lowest, 100 is highest and have values all throughout).

But I assume I could divide all variable by 100 and use that technique?

Re: Regression Technique

Ksharp — Thu, 09 Mar 2017 04:23:15 GMT

I could try GAMMA distribution.
Check the example of GENMOD.

Example 44.3: Gamma Distribution Applied to Life Data